## wpiea2025187-source-pdf

## Source details

**Canonical URL:** [wpiea2025187-source-pdf](https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025187-source-pdf.pdf)

## Other formats

- [Markdown version](/-/media/files/publications/wp/2025/english/wpiea2025187-source-pdf.pdf.md)
- [Structured JSON version](/-/media/files/publications/wp/2025/english/wpiea2025187-source-pdf.pdf.json)

---

### Executive summary — Key context and motivation
- Türkiye’s labor force participation (LFP) is "at around 53 percent" and "considerably lower than in peer economies."
- Shortfall concentrated in secondary earners (mostly women); Türkiye’s female LFP rate is much lower than male LFP and female LFP in other countries. The gap between female and male LFP in Türkiye is among the largest in the G20.
- Closing LFP gaps could generate "sizeable labor supply and GDP gains."
- Childcare affordability is a significant constraint, especially for low-income households:
  - "Monthly childcare costs: TL 2600 (as of 2022)"
  - Attendance of pre-school or childcare facilities is widespread in the top-income decile but very limited in the bottom-income decile.

### Policy channels analyzed
- Two fiscal channels analyzed:
  - Removing tax disadvantages faced by Turkish secondary earners (2022 Minimum Living Allowance reform).
  - Introducing conditional childcare subsidies to reduce childcare spending and boost caregivers’ disposable income.

### Models and toolkit used
- Four models built for Türkiye:
  - IMF’s TaxFit microsimulation model (TaxFit) to compute household net income given taxes, social contributions, benefits and regulations.
    - Net income formula used in the model: netinc = gross + sa + fb + care + ub + iw − it − ssc
  - Heckman selection model to simulate potential earnings for non-working individuals (selection equation + modified Mincer earnings equation).
  - Two-way fixed effects econometric model to estimate the impact of the MLA reform on female employment (married women as treatment, single women as control; variations: all working-age women; women with no minor children; women with minor children).
  - Discrete choice model to estimate labor supply response to changes in net income.
- Data sources: Turkstat Labor Force micro data and country-specific laws/regulations (Income Tax Law no. 193; Social Insurance and General Health Insurance Law no. 5510; Labor Law no. 4857; Unemployment Insurance Law no. 4447).

### Findings — Minimum Living Allowance (MLA) 2022 reform
- Background:
  - Türkiye’s MLA system (Asgari Geçim İndirimi) previously exacerbated labor supply disadvantages for married secondary earners (typically women).
  - January 2022 reform: MLA abolished and replaced with a PIT exemption of all earnings up to the minimum wage, regardless of family structure.
- Pre-reform MLA mechanics (preserved percentages):
  - Allowance structure: 50% for the taxpayer; 10% for a non-working spouse; 7.5% for each of the first two children; 10% for a third child; 5% for each additional child (up to a maximum total allowance of 85% of the minimum wage).
  - Allowance granted on an individual basis, but only one spouse could claim the allowance related to children.
- Distributional and behavioral implications:
  - Reform increased the basic non-taxable allowance from between 50 and 85 percent of the minimum wage (depending on household composition) pre-reform, to 100 percent of the minimum wage post-reform, making the tax system more progressive.
  - Although equal for men and women, larger effects expected among women because their labor supply is more elastic to changes in net incomes.
- Estimated impacts:
  - "The reduction in effective tax rates induced by the 2022 reform did lead to an increase in total employment, driven by secondary earners, who are mostly women."
  - Post-reform outcomes for women included: "the probability of women working formally increased, while the probability that they chose to stay at home or work informally declined post reform."
- Estimated elasticities (extensive margin):
  - Removal of the Dependent Spouse Allowance: 0.04-0.15.
  - Wider MLA reform: 0.21-0.40.
  - Comparison benchmarks: average extensive margin elasticity of 0.25 (Chetty et al., 2013); surveyed elasticities between 0.5 and 1.0 for women and low-education individuals (Immervoll et al., 2007).
- Aggregate labor supply impacts (model results):
  - Female labor force participation and employment increased by 3 percentage points.
  - Male labor force participation and employment increased by 1.5 percentage points.
  - Implied total labor force increase of around 2.5 percentage points.

### Two-way fixed effects model (specification and measurement)
- Model specification:
  - y_it = β0 + β1 ⋅ treatment_i + β2 ⋅ time_t + β3 ⋅ (time_t ⋅ θ_i^DSA) + β4 ⋅ (time_t ⋅ θ_i^NTA) + ε_it
- Definitions:
  - y_it: binary indicator of whether a woman is working.
  - treatment_i: identifies married women (beneficiaries of dependent spouse allowance pre-reform) vs single women.
  - time_t: zero in 2021 and one in 2022.
  - θ_i^DSA: change in the tax burden resulting from the removal of the dependent spouse allowance.
  - θ_i^NTA: change in the overall tax burden excluding the dependent spouse allowance.
  - ε_it: idiosyncratic errors clustered by cohort status.
- Measurement:
  - θ_i^DSA and θ_i^NTA are reported in percentage point changes in individuals’ effective net tax rates (negative values reflect reductions in tax-induced work disadvantages).
- Parameters of interest:
  - β3 and β4 measure expected change in employment between treatment and control groups pre- and post-reform.

### Discrete choice model (purpose and mechanics)
- Purpose:
  - Assess impact of the reform on probabilities of not working, working formally, or working informally for each woman.
- Utility specification:
  - U_itj = V_itj + ε_itj where V_itj = x_itj β
  - j choices: not working, working formally, working informally.
  - x_itj includes individual-specific characteristics (education, number of children, region) and choice-dependent variables (taxes, benefits, income levels).
- Choice assumptions and drivers (Table A preserved):
  - Work formally: Income = Yes; Benefits = No; Costs = Yes
  - Work informally: Income = Yes; Benefits = Yes; Costs = No
  - Not work: Income = No; Benefits = No; Costs = No
- Estimation and prediction:
  - Conditional logit used to predict probabilities before and after the reform.
  - Counterfactual incomes for non-working individuals predicted with Heckman model; taxes and benefits calculated with TaxFit.
- Key assumptions:
  - Formal workers pay taxes and forgo benefits; informal workers receive benefits if earnings low enough to qualify.
  - Benefits in practice targeted to very low-income families; LFS data 2021-2022 show few formally-employed households qualify.

### Data (coverage and sample adjustments)
- Turkstat Labor Force micro data (LFS).
- Raw dataset: nearly 100 variables for around 360 thousand working-age individuals in 2021–2022.
- Sample adjusted for TaxFit specifications: excludes grandparents, unrelated roommates, children over 25, households with multiple spouses, and households with no identified household head.
- Adjusted dataset:
  - Includes observations for 250 thousand working-age individuals in each year (representing around 33.5 million individuals).
  - Gender composition: 49 percent men and 51 percent women.
  - Labor market status shares:
    - 37 percent are outside of the labor force.
    - 5 percent are unemployed.
    - 43 percent are employed formally.
    - 14 percent are employed informally.
  - Among working-age women in the sample:
    - Around one third were working: 22 percent in the formal sector and 10 percent in the informal sector.
    - 4 percent were unemployed.
    - 61 percent were outside of the labor force.
  - Approximately 82 percent of working-age women were married.
  - Around 30 percent had children under the age of 5.
- Working age definition:
  - 16–58 years for women and 16–60 years for men.

### Childcare subsidy policies analyzed (design and variants)
- Policy objective:
  - Subsidize childcare services to promote LFP, particularly for low-income households and secondary earners (typically women).
- Baseline subsidy proposal:
  - Full direct universal subsidy of the total cost of private childcare services for households where both parents work.
  - Assumed TL 2,600 per child monthly for each eligible child (median private childcare cost as of 2022).
  - Administered as a household benefit (could also be administered to providers).
- Five subsidy reform variants:
  - Baseline reform: Full direct universal cash subsidy for each child age 0 to 5 (exclusive) conditional on both parents working (formally or informally) and the child being in childcare.
  - Formal employment: Baseline with eligibility limited to households where all parents are engaged in formal employment.
  - Two-child maximum: Baseline with an additional limit of two eligible children per household.
  - Income cap: Baseline with additional eligibility limit to households with annual earnings of TL120,000 or less (75th percentile).
  - Ages below 3 years: Baseline limited to children aged between 0 and 3 (exclusive).
  - Reference scenario 1: Unconditional direct transfer (ages below 5 years): cash transfer for each child aged 0 to 5 (exclusive) not conditional on employment or earnings.
- Alternative policy simulations (Annex II):
  - Half subsidy; Expense deduction; Non-refundable tax credit; Refundable tax credit; Direct transfer (ages 0–5).

### Childcare subsidy simulation process
- Three-step simulation:
  - Step 1. Estimate deadweight cost: difference in net taxes payable by households where all parents worked prior to reform.
    - Deadweight Cost = Σ (τ_i_post−reform − τ_i_pre−reform) for i ∈ Π, where Π is set of households where all adults work.
  - Step 2. Predict employment probability changes for individuals affected by childcare policy:
    - Use two-way fixed effects model; predicted probability of formal work equals:
      - 1 for anyone observed working;
      - 0 for non-working individuals not eligible for childcare subsidies;
      - between 0 and 1 for eligible non-working individuals.
    - Calculation multiplies percent change in net income due to childcare policy by elasticity of employment with respect to change in METR for the dependent spouse allowance estimated in section D.
  - Step 3. Estimate fiscal costs:
    - Use TaxFit model to simulate social assistance contributions, family benefits, and income taxes pre-reform; simulate taxes and benefits assuming employment post-reform.
    - Fiscal Cost = Deadweight Cost + Σ ρ_i · (Δsa_i + Δfb_i + Δit_i)
      - ρ_i is change in individual i’s employment probability post reform.
      - Δsa_i, Δfb_i, Δit_i are changes in social assistance, family benefits, and income taxes assuming i becomes employed post reform.
    - Note: social security contributions assumed earmarked and not improving fiscal balance.

### Childcare subsidy simulation results and fiscal costs
- LFP impacts:
  - Conditional childcare subsidies yield "a 4 percentage point increase in total LFP in Türkiye" (Executive Summary statement).
  - Targeted cohort (working-age parents with young children):
    - LFP gender gap decreases from around 56pp to roughly 23pp under the baseline reform.
  - Entire population:
    - LFP gender gap decreases from around 36 to roughly 30 percentage points.
  - Overall total LFP:
    - Pre-reform total LFP 53 percent.
    - Baseline reform total LFP 57 percent.
    - Change driven by an 8pp increase in female LFP and a 1pp increase in male participation.
- Coverage and affected children:
  - Estimated number of affected children ranges from 3mn (Baseline reform) to 1mn (Ages 0-3 reform).
- Fiscal cost estimates:
  - Fiscal cost estimates range "between 0.2 and 0.6 percent of GDP."
  - Highest-impact subsidy costs 0.6 percent of GDP.
  - Targeted policies cost between 0.2 and 0.6 percent of GDP.
  - All estimated costs are below OECD average annual spending on early childhood care and education of 0.8 percent of GDP.
  - Türkiye’s current annual spending on early childhood care and education is around 0.3 percent of GDP.
  - Estimated half of total costs are deadweight costs and half variable costs; both only partly offset by higher social security contributions from increased employment.
- Conditional subsidies vs unconditional direct transfers:
  - Unconditional direct transfer counterfactual: TL 2,600 per month per child aged 0–5 regardless of parental employment; represents windfall gains with no LFP effect and potentially adverse participation effects via income effect.
  - Conditional subsidies:
    - More effective at increasing LFP.
    - Lower net fiscal cost because as secondary earners start to work, fiscal spending on social assistance falls and PIT revenues rise.
  - Policy implication: making transfers conditional on parents’ employment and children’s childcare attendance could both support welfare and increase LFP while reducing fiscal costs.

### Annex I — Estimation results (selected)
- Two-way Fixed Effects Model — regression highlights (Women; Men):
  - Women (N: 133694; 29284; 104410; RMSE: 0.4634; 0.4843; 0.4591):
    - Dependent Spouse Allowance: -1.8976*** (0.2912); -0.7331 (0.7607); -1.9214*** (0.2993)
    - Non-taxable allowance: -5.3064*** (0.4478); -5.2059*** (0.7352); -5.3126*** (0.5495)
  - Men (N: 52874; 22144; 30730; RMSE: 0.3804; 0.4435; 0.361):
    - Dependent Spouse Allowance: 1.0476** (0.5165); 0.8098 (0.6985); 0.8789 (0.5573)
    - Non-taxable allowance: 1.9449*** (0.3076); 1.9317*** (0.6218); 1.7039*** (0.3240)
  - Note: Standard errors in parenthesis.
- Discrete Choice Model — predictive margins (Number of obs = 48,333,057):
  - 2021 margins:
    - Not work: 0.56; Work formal: 0.27; Work informal: 0.18
  - 2022 margins:
    - Not work: 0.54; Work formal: 0.29; Work informal: 0.17
- Mean percentage point change in individuals’ effective net tax rates from the MLA reform (Table 3):
  - All women: Theta DSA −2.50; Theta NTA (married) −2.25; Theta NTA (single) −2.47
  - Without children: −2.24; −3.20; −3.54
  - With children: −2.53; −2.15; −1.73

### Overall implications and policy relevance
- Fiscal policy can boost LFP in Türkiye via:
  - Tax policy adjustments that remove implicit tax disincentives for secondary earners (as in the 2022 MLA reform).
  - Expenditure policy (conditional childcare subsidies) that reduces private childcare spending and raises after-tax returns to work for caregivers.
- The combination of removing PIT disadvantages (as implemented in 2022) and introducing conditional childcare subsidies is estimated to materially increase employment among secondary earners (mostly women) at fiscally feasible costs.

*Tax Reform, Subsidies, and Labor Supply: Lessons from Türkiye — Working Paper No. WP/2025/187*

### Executive Summary ......................................................................................................

### Executive Summary

### A. Key context and motivation
- Türkiye’s labor force participation (LFP) is "at around 53 percent" and "considerably lower than in peer economies."
- The shortfall is concentrated in secondary earners (mostly women); Türkiye’s female LFP rate is much lower than male LFP and female LFP in other countries. The gap between female and male LFP in Türkiye is among the largest in the G20.
- Closing LFP gaps could generate "sizeable labor supply and GDP gains."
- Childcare affordability is a significant constraint, especially for low-income households:
  - "Monthly childcare costs: TL 2600 (as of 2022)"
  - Attendance of pre-school or childcare facilities is widespread in the top-income decile but very limited in the bottom-income decile.

### B. Policy channels analyzed
- Two fiscal channels analyzed:
  - Removing tax disadvantages faced by Turkish secondary earners (2022 Minimum Living Allowance reform).
  - Introducing conditional childcare subsidies to reduce childcare spending and boost caregivers’ disposable income.

### C. Models and toolkit used
- Four models built for Türkiye:
  - IMF’s TaxFit microsimulation model (TaxFit) to compute household net income given taxes, social contributions, benefits and regulations.
    - Net income formula used in the model: netinc = gross + sa + fb + care + ub + iw − it − ssc
  - Heckman selection model to simulate potential earnings for non-working individuals (selection equation + modified Mincer earnings equation).
  - Two-way fixed effects econometric model to estimate the impact of the MLA reform on female employment (married women as treatment, single women as control; variations: all working-age women; women with no minor children; women with minor children).
  - Discrete choice model to estimate labor supply response to changes in net income.
- Data sources: Turkstat Labor Force micro data and country-specific laws/regulations (Income Tax Law no. 193; Social Insurance and General Health Insurance Law no. 5510; Labor Law no. 4857; Unemployment Insurance Law no. 4447).

### D. Findings — Minimum Living Allowance (MLA) 2022 reform
- Background:
  - Türkiye’s MLA system (Asgari Geçim İndirimi) previously exacerbated labor supply disadvantages for married secondary earners (typically women).
  - A 2022 reform "improved work incentives for all earners, but especially for married secondary earners, by enhancing the progressivity of the tax system and removing family-based tax provisions."
- Estimated impacts:
  - "The reduction in effective tax rates induced by the 2022 reform did lead to an increase in total employment, driven by secondary earners, who are mostly women."
  - Post-reform outcomes for women included: "the probability of women working formally increased, while the probability that they chose to stay at home or work informally declined post reform."

### E. Findings — Conditional childcare subsidies
- Policy simulated: introduction of conditional childcare subsidies (targeted to affect caregivers’ net disposable income and employment incentives).
- Simulation approach:
  - Stepwise: estimate deadweight cost of the childcare policy; predict employment probability changes for affected individuals; estimate fiscal costs of the childcare policy.
- Estimated impacts and fiscal costs:
  - Conditional childcare subsidies yield "a 4 percentage point increase in total LFP in Türkiye."
  - Fiscal cost estimates range "between 0.2 and 0.6 percent of GDP."
  - "Conditional subsidies achieve better labor market outcomes and further reduce fiscal costs compared to direct transfers."

### F. Overall implications and policy relevance
- Fiscal policy can play an important role in boosting LFP in Türkiye via:
  - Tax policy adjustments that remove implicit tax disincentives for secondary earners.
  - Expenditure policy (childcare subsidies) that reduces private childcare spending and raises after-tax returns to work for caregivers.
- The combination of removing PIT disadvantages (as implemented in 2022) and introducing conditional childcare subsidies is estimated to materially increase employment among secondary earners (mostly women) at fiscally feasible costs.

_Executive Summary — wpiea2025187-source-pdf_

### 2022. This  estimated  model  was  then  used  again—in  section  E  of  this  paper—to  predict the labor  supply

### wpiea2025187-source-pdf - 2022. This  estimated  model  was  then  used  again—in  section  E  of  this  paper—to  predict the labor  supply

### Two-way fixed effects model
- Model specification:
  - y_it = β0 + β1 ⋅ treatment_i + β2 ⋅ time_t + β3 ⋅ (time_t ⋅ θ_i^DSA) + β4 ⋅ (time_t ⋅ θ_i^NTA) + ε_it
- Definitions and identification:
  - y_it represents a binary indicator of whether a woman is working or not.
  - treatment_i identifies if a woman is in the treatment group that benefits from the dependent spouse allowance—because she is married rather than single.
  - time_t equals zero in 2021 and one in 2022.
  - θ_i^DSA is the change in the tax burden resulting from the removal of the dependent spouse allowance.
  - θ_i^NTA is the change in the overall tax burden excluding the dependent spouse allowance.
  - ε_it are independent and uncorrelated idiosyncratic errors, clustered by cohort status defined by age group, gender, family status, education, geographic region, and earnings decile of the spouse.
- Measurement:
  - Both tax burden indicators (θ_i^DSA and θ_i^NTA) are reported in terms of percentage point changes in an individuals’ effective net tax rates.
  - These indicators are negative, reflecting the reduction in tax-induced work disadvantages following the reform.
- Parameters of interest:
  - The coefficients of interest correspond to the interaction terms between time and the tax burden variables (β3 and β4), which measure the expected change in employment between the treatment and control groups pre- and post-reform.

### Discrete Choice Model (Section B.4)
- Purpose:
  - To assess the impact of the reform on informal work by estimating probabilities of not working, working formally, or working informally for each woman.
- Theoretical basis:
  - Random utility framework (McFadden 2001; Ben-Akiva & Lerman, 1985; Chen et al., 2016).
- Utility specification:
  - U_itj = V_itj + ε_itj where V_itj = x_itj β
  - j denotes choices (not working, working formally, working informally).
  - V_itj is a function of regressors x_itj including individual-specific characteristics (education, number of children, region of residence) and choice-dependent variables (taxes, benefits, income levels).
  - β is the vector of coefficients associated with variables in x_itj.
- Choice assumptions and drivers (Table A):
  - Work formally: Income = Yes; Benefits = No; Costs = Yes
  - Work informally: Income = Yes; Benefits = Yes; Costs = No
  - Not work: Income = No; Benefits = No; Costs = No
- Estimation and prediction:
  - Estimated utilities for each labor market choice are used to predict probabilities of not working, working formally, or working informally before and after the reform using a conditional logit framework (McFadden, 1972).
- Data and simulation inputs:
  - Labor market outcomes observed in the 2021-2022 Turkstat LFS household surveys.
  - Tax liabilities and benefit entitlements for those who work (formally or informally) simulated with the TaxFit microsimulation model.
  - For each non-working individual, counterfactual incomes are predicted using the Heckman model, and associated taxes and benefits calculated with TaxFit.
- Modeling assumptions and caveats:
  - Assumed formal workers pay taxes and forgo benefits, informal workers receive benefits.
  - In practice, benefits are targeted to very low-income families regardless of formal or informal employment status; LFS data for 2021-2022 show few formally-employed households qualify, making benefit receipt predominantly observed among informal workers.
  - Formal workers lose eligibility for benefits upon employment if they previously met the income criteria; informal workers receive benefits only if their earnings remain low enough to qualify.
- Coefficient structure:
  - Coefficients are constant across choices for variables that vary across choices (such as net tax rates and incomes).
  - The coefficient on choice-invariant variables (such as socio-demographic characteristics) is heterogeneous across choices.

### Data (Section C)
- Source and coverage:
  - Turkstat’s Labor Force micro data (LFS).
  - Raw dataset covers nearly 100 variables for around 360 thousand working-age individuals in 2021–2022.
- Sample adjustments:
  - TaxFit only simulates taxes and benefits for working-age single-mothers, single-fathers, and nuclear-family households (parents and their dependent children).
  - LFS data for grandparents, unrelated roommates, children over 25 years old, households with multiple spouses and those with no identified household head were excluded.
- Adjusted dataset:
  - Includes observations for 250 thousand working-age individuals in each year (representing around 33.5 million individuals).
  - Gender composition: 49 percent men and 51 percent women.
  - Labor market status shares:
    - 37 percent are outside of the labor force.
    - 5 percent are unemployed.
    - 43 percent are employed formally.
    - 14 percent are employed informally.
  - Among working-age women in the sample:
    - Around one third were working: 22 percent in the formal sector and 10 percent in the informal sector.
    - 4 percent were unemployed.
    - 61 percent were outside of the labor force.
- Demographic and labor-market patterns:
  - Approximately 82 percent of working-age women were married.
  - Around 30 percent had children under the age of 5.
  - Prior to the reform:
    - Employment rate for single women stood at around 45 percent.
    - Employment rates for married women ranged from 20 percent (for women with children and a spouse working informally) to 40 percent (for women without children and a spouse working formally).
  - Post reform: employment rates increased across the board for women (Figure 6 referenced).
  - Secondary earners:
    - Most secondary earners are women.
    - Men are the primary earner in around 80 percent of households with at least one person working and in around 60 percent of two-earner households (Figure 7 referenced).
- Working age definition:
  - Working age defined as 16–58 years for women and 16–60 years for men.

### Case 1: Minimum Living Allowance Reform (Section D, D1)
- Reform description:
  - Until recently, Türkiye’s Minimum Living Allowance (MLA) system granted partial tax allowances on personal income—up to the minimum wage—for all workers and additional tax allowances depending on family structure (spouse’s working status and number of children).
  - In January 2022, the Minimum Living Allowance (MLA) was abolished and replaced with a PIT exemption of all earnings up to the minimum wage, regardless of family structure.
- Mechanics of pre-reform MLA (footnote detail preserved):
  - The MLA system provided a tax allowance as a percentage of the minimum wage tax obligations: 50% for the taxpayer; 10% for a non-working spouse; 7.5% for each of the first two children; 10% for a third child; 5% for each additional child (up to a maximum total allowance of 85% of the minimum wage).
  - The allowance was granted on an individual rather than household basis, but only one spouse could claim the allowance related to children.
- Effects and distributional implications:
  - The reform increased the basic non-taxable allowance from between 50 and 85 percent of the minimum wage (depending on household composition) pre-reform, to 100 percent of the minimum wage post-reform, making the tax system more progressive.
  - Although equal for men and women, the change could have larger effects among women because their labor supply is more elastic to changes in net incomes (Evers et. al 2008), reflecting women’s greater propensity to take on a larger share of home duties (Alesina et. al, 2011).
  - Women tend to be over-represented at the bottom of the income distribution; a more progressive tax system can be associated with increases in female labor supply (Coelho et al, 2024).
  - The removal of the PIT allowance for non-working spouses (“dependent spouse allowance”) reduced work disadvantages for secondary earners, primarily women.
  - Prior to the reform, the household’s net tax allowance would increase by 40 percent of the tax payable by a minimum wage worker (footnote context).
- Notes on pre-reform basic allowance (footnote detail preserved):
  - Prior to the MLA reform, workers without a dependent spouse or dependent children would benefit from a basic non-taxable allowance equal to 50 percent of the minimum wage, while workers with dependent children would receive a larger allowance up to 85 percent of the minimum wage, depending on the number of dependent children.

*IMF WORKING PAPERS Tax Reform, Subsidies, and Labor Supply: Lessons from Türkiye — INTERNATIONAL MONETARY FUND*

### 58. The employment rate is the ratio of working

### 58. The employment rate is the ratio of working-age employed to the working age population. Sample adjusted to Taxfit specifications.

### Tax Allowance Reform and Labor Supply
- Pre- and post-reform tax allowance structure (example households without children):
  - (A) Non-working unmarried woman: Pre-reform 0, Post-Reform 0
  - (B) Working unmarried woman: Pre-reform 50, Post-Reform 100
    - (B)-(A) = Tax allowance gain from working (unmarried woman), p.p. 50, 100
  - (C) Working married woman with non-working spouse: Pre-reform 50 + 10 = 60, Post-Reform 100
  - (D) Working married woman with working spouse: Pre-reform 50 + 50 = 100, Post-Reform 100 + 100 = 200
    - (D)-(C) = Tax allowance gain from working (married woman), p.p. 40, 100
- Reform detail:
  - Before reform: an unmarried working woman received a 50 percent allowance; a married working woman could receive 50 + 10 = 60 percent (or 50 + 50 = 100 percent depending on spouse status) as examples.
  - After the reform, the tax allowance from working was increased to 100pp for married and unmarried women, removing the discrepancy.

### Estimating Labor Supply Response to the MLA Reform
- Four-model estimation approach; multi-step process (Figure 8):
  - Step 1. Generate simulated gross income for non-working individuals using the Heckman model.
    - Simulated gross income of non-working individuals is lower than observed gross income of working individuals by about 15 percent on average.
    - Women have lower gross income than men in both observed and simulated data.
  - Step 2. Generate simulated net income for working and non-working individuals using the TaxFit model and household characteristics; simulate net income under working and non-working status holding other household characteristics constant.
  - Step 3. Compute each individual's marginal effective tax rate (METR) from entering employment:
    - METR_i = (net taxes_i_working − net taxes_i_not working) / (earnings_i_working − earnings_i_not working) = (Δtaxes_i / Δearnings_i)
    - METR calculated pre- and post-reform. Post-reform scenario uses 2022 tax code and benefit policies including new policies that replaced the MLA system; pre-reform scenario uses 2022 code adjusted to include 2021 MLA provisions so the MLA reform is isolated as the only source of differences in METRs between 2021 and 2022.
    - After the reform, the METR from entering work for married women decreased by more than for unmarried women.
  - Step 4. Estimate labor supply response with and without MLA reform using two-way fixed effects and discrete choice models.
- Model and elasticity findings:
  - Two-way fixed effects model estimated elasticities that are comparable to other studies.
  - Results indicate:
    - (i) Female labor supply was responsive to the reduction in effective tax rates caused by the increase of the basic non-taxable allowance (and more responsive than men’s).
    - (ii) Women’s labor supply was also responsive to the removal of the dependent spouse allowance, though to a lesser degree than the larger reform to the basic allowance.
  - Estimated elasticities referenced:
    - Removal of the Dependent Spouse Allowance: 0.04-0.15.
    - Wider MLA reform: 0.21-0.40.
    - These are slightly higher than the average extensive margin elasticity of 0.25 surveyed by Chetty et al. (2013), but smaller than surveyed elasticities of between 0.5 and 1.0 for women and low-education individuals surveyed by Immervoll et al. (2007).
  - Sensitivity: using simulated incomes two standard errors higher or lower does not alter magnitude or significance of estimations.
- Aggregate labor supply impacts (model results):
  - As a result of the overall MLA reform:
    - Female labor force participation and employment increased by 3 percentage points.
    - Male labor force participation and employment increased by 1.5 percentage points.
    - Implied total labor force increase of around 2.5 percentage points.
  - Discrete choice model results broadly consistent:
    - MLA reform increased women’s probability of working in the formal sector and lowered probability of staying at home or working informally.

### Childcare Subsidy Policies (Case 2)
- Policy objective:
  - Introduce a subsidy on childcare services to promote labor force participation (LFP), particularly for low-income households and secondary earners (typically women).
- Baseline subsidy proposal:
  - Full direct universal subsidy of the total cost of private childcare services for households where both parents work.
  - Assumed TL 2,600 per child monthly for each eligible child, in line with estimated median costs for private childcare services as of 2022.
  - Administered as a household benefit (could also be administered to providers).
  - Simulations note this may be an upper bound given public childcare is typically cheaper and price differentials may reflect existing subsidies.
- Five subsidy reform variants (Table D):
  - Baseline reform: Full direct universal cash subsidy for each child age 0 to 5 (exclusive) conditional on both parents working (formally or informally) and the child being in childcare.
  - Formal employment: Baseline with eligibility limited to households where all parents are engaged in formal employment.
  - Two-child maximum: Baseline with additional limit of two eligible children per household.
  - Income cap: Baseline with additional eligibility limit to households with annual earnings of TL120,000 or less (75th percentile).
  - Ages below 3 years: Baseline limited to children aged between 0 and 3 (exclusive).
  - Reference scenario 1: Unconditional direct transfer (ages below 5 years): cash transfer for each child aged 0 to 5 (exclusive) not conditional on employment or earnings.

### Childcare Subsidy Simulation Process
- Three-step simulation (Figure 13):
  - Step 1. Estimate deadweight cost: difference in net taxes payable by households where all parents worked prior to reform.
    - Deadweight Cost = Σ (τ_i_post−reform − τ_i_pre−reform) for i ∈ Π, where Π is set of households where all adults work.
  - Step 2. Predict employment probability changes for individuals affected by childcare policy:
    - Use two-way fixed effects model; predicted probability of formal work equals:
      - 1 for anyone observed working,
      - 0 for non-working individuals not eligible for childcare subsidies,
      - between 0 and 1 for eligible non-working individuals.
    - Calculation multiplies percent change in net income due to childcare policy by elasticity of employment with respect to change in METR for the dependent spouse allowance estimated in section D.
  - Step 3. Estimate fiscal costs:
    - Use TaxFit model to simulate social assistance contributions, family benefits, and income taxes pre-reform; simulate taxes and benefits assuming employment post-reform.
    - Fiscal Cost = Deadweight Cost + Σ ρ_i · (Δsa_i + Δfb_i + Δit_i)
      - ρ_i is change in individual i’s employment probability post reform.
      - Δsa_i, Δfb_i, Δit_i are changes in social assistance, family benefits, and income taxes assuming i becomes employed post reform.
    - Note: social security contributions assumed earmarked and not improving fiscal balance.

### Childcare Subsidy Simulation Results
- LFP impacts:
  - Introduction of conditional childcare subsidy could lead to a large increase in LFP.
  - For targeted cohort (working-age parents with young children):
    - LFP gender gap decreases from around 56pp to roughly 23pp under the baseline reform.
  - For the entire population (including non-targeted individuals):
    - LFP gender gap decreases from around 36 to roughly 30 percentage points.
  - Overall total LFP:
    - Pre-reform total LFP 53 percent.
    - Baseline reform total LFP 57 percent.
    - Change driven by an 8pp increase in female LFP and a 1pp increase in male participation.
- Coverage and affected children:
  - Estimated number of affected children ranges from 3mn (Baseline reform) to 1mn (Ages 0-3 reform).
- Fiscal costs (Figure 15 and discussion):
  - Highest-impact subsidy costs 0.6 percent of GDP.
  - Targeted policies cost between 0.2 and 0.6 percent of GDP.
  - All estimated costs are below OECD average annual spending on early childhood care and education of 0.8 percent of GDP.
  - Türkiye’s current annual spending on early childhood care and education is around 0.3 percent of GDP.
  - Estimated half of total costs are deadweight costs and half variable costs; both only partly offset by higher social security contributions from increased employment.
- Conditional subsidies vs unconditional direct transfers:
  - Unconditional direct transfer counterfactual: TL 2,600 per month per child aged 0–5 regardless of parental employment; represents windfall gains with no LFP effect and potentially adverse participation effects via income effect.
  - Conditional subsidies:
    - More effective at increasing LFP.
    - Lower net fiscal cost because as secondary earners start to work, fiscal spending on social assistance falls and PIT revenues rise.
  - Empirical positioning:
    - Conditional subsidies cost less and are more effective than unconditional direct transfers.
    - Türkiye’s existing family support programs (example: childbirth support program amounts provided in variable lump sums and monthly amounts) are unconditional; making transfers conditional on parents’ employment and children’s childcare attendance could both support welfare and increase LFP while reducing fiscal costs.

### Conclusion
- Empirical evidence from Türkiye indicates:
  - A 2022 reform that removed tax disadvantages for secondary earners and enhanced tax progressivity boosted total labor supply.
  - Conditional childcare subsidies can lead to a large increase in labor force participation at relatively low fiscal costs.

*Source: Turkish Statistical Institute Labour Force Statistics Micro Data, 2021 and 2022, and IMF staff calculations; IMF Working Papers — Tax Reform, Subsidies, and Labor Supply: Lessons from Türkiye.*

### Annex I. Estimation Results

### Annex I. Estimation Results

### Two-way Fixed Effects Model — Regression Results
- Women (Dependent variable columns: Single v Married; Single v Married (no children); Single v Married (with children))
  - Dependent Spouse Allowance: -1.8976*** (0.2912); -0.7331 (0.7607); -1.9214*** (0.2993)
  - Non-taxable allowance: -5.3064*** (0.4478); -5.2059*** (0.7352); -5.3126*** (0.5495)
  - Time FEs: yes; Treatment FEs: yes
  - N: 133694; 29284; 104410
  - RMSE: 0.4634; 0.4843; 0.4591

- Men (Dependent variable columns: Single v Married; Single v Married (no children); Single v Married (with children))
  - Dependent Spouse Allowance: 1.0476** (0.5165); 0.8098 (0.6985); 0.8789 (0.5573)
  - Non-taxable allowance: 1.9449*** (0.3076); 1.9317*** (0.6218); 1.7039*** (0.3240)
  - Time FEs: yes; Treatment FEs: yes
  - N: 52874; 22144; 30730
  - RMSE: 0.3804; 0.4435; 0.361

- Note: Standard errors in parenthesis.

### Discrete Choice Model — Predictive Margins (Number of obs = 48,333,057)
- 2021 predictive margins (Outcome — Margin; Delta-method Margin std. err.; z; P>|z|; [95% conf. interval])
  - Not work: 0.56; 0.00; 4502.37; 0.00; 0.56; 0.56
  - Work formal: 0.27; 0.00; 2434.06; 0.00; 0.27; 0.27
  - Work informal: 0.18; 0.00; 1859.32; 0.00; 0.18; 0.18

- 2022 predictive margins (Outcome — Margin; Delta-method Margin std. err.; z; P>|z|; [95% conf. interval])
  - Not work: 0.54; 0.00; 4400.1; 0.00; 0.54; 0.55
  - Work formal: 0.29; 0.00; 2550.72; 0.00; 0.29; 0.29
  - Work informal: 0.17; 0.00; 1810.27; 0.00; 0.17; 0.17

### Mean percentage point change in individuals’ effective net tax rates from the MLA reform (Table 3)
- Sample Theta DSA; Theta NTA (married); Theta NTA (single)
  - All women: -2.50; -2.25; -2.47
  - Without children: -2.24; -3.20; -3.54
  - With children: -2.53; -2.15; -1.73

### Figures and Notes
- Figure A: Two-way Fixed Effects Model Results — Labor force participation elasticity, men
  - Note: Figure does not include time and treatment fixed effects.
  - Source: Turkstat Household Labor Force Survey, IMF Taxfit model, and IMF staff calculations.

---

### Annex II. Alternative Childcare Policy Simulations

### Policy descriptions (as simulated)
- Baseline reform
  - Full cash subsidy is provided for each child of childcare age (0 to 5) to households where all parents are working.
- Half subsidy
  - Baseline scenario, with 50% of childcare expenses covered.
- Expense deduction
  - The full amount of childcare expenses can be claimed as deductible expenses from personal income tax payable for each child of childcare age (0 to 5) to households where all parents are working.
- Non-refundable tax credit
  - A non-refundable (wastable) tax credit equal to the full amount of childcare expenses is applied to income tax payable for each child of childcare age (0 to 5) to households where all parents are working.
- Refundable tax credit
  - A refundable (non-wastable) tax credit equal to the full amount of childcare expenses is applied to income tax payable for each child of childcare age (0 to 5) to households where all parents are working.
- Direct transfer (ages 0–5)
  - Reference scenario: A direct cash transfer is paid to families for each child aged between 0 and 5 (inclusive). There are no employment or earnings eligibility requirements.

### Simulated outcomes (figures)
- Figure A: Simulated LFP gender gap pre- and post-reforms (male minus female LFP)
- Figure B: Simulated fiscal costs of proposed reforms

- Note: Direct transfers signify direct cash payments to children of childcare age of TL2600 per month (in 2022 terms). Early childhood educational development services (ECED) focus on children aged under 3 years.
- Source: Turkstat Household Labor Force Survey, IMF Taxfit model, and IMF staff calculations.

*Tax Reform, Subsidies, and Labor Supply: Lessons from Türkiye — Working Paper No. WP/2025/187*

---


_Source: https://www.imf.org/-/media/files/publications/wp/2025/english/wpiea2025187-source-pdf.pdf_